1. Field of the Invention
This invention relates to an oxygen doping method into a gallium nitride crystal and an oxygen-doped n-type gallium nitride single crystal substrate for producing light emitting diodes (LEDs), laser diodes (LDs) or other electronic devices of groups 3 and 5 nitride semiconductors. Nitride semiconductors means GaN, InGaN, InAlGaN and so on which are grown as thin films on a sapphire substrate. An activation layer is a GaInN layer. Other parts are mainly GaN layers. Thus, the light emitting diodes based upon the nitride semiconductors are represented as GaN-LEDs or InGaN-LEDs which mean the same LEDs.
This application claims the priority of Japanese Patent Application No. 2001-113872 filed on Apr. 12, 2001 which is incorporated herein by reference.
2. Description of Related Art
Light emitting devices making use of nitride semiconductors have been put on the market as blue-light LEDs. At present, all of the available nitride light emitting devices are made upon sapphire substrates. An epitaxial wafer is obtained by growing a GaN film, a GaInN film and so forth upon a C-plane single crystal sapphire substrate heteroepitaxially. A unique n-dopant for GaN, AlInGaN, or InGaN thin films is silicon (Si). Silicon acts as an n-impurity in GaN by replacing a gallium site. A series of wafer processes produces GaInN-LEDs on the on-sapphire epitaxial wafer. A lattice constant of sapphire (α-Al2O3) is different from that of gallium nitride. Despite the large lattice misfit, a sapphire substrate allows gallium nitride to grow heteroepitaxially on it. The on-sapphire GaN includes great many dislocations. In spite of the many dislocations, the GaN films on sapphire are stable and endurable.
Sapphire is a crystal of a trigonal symmetry group. C-plane of sapphire has quasi-three fold rotation symmetry. Gallium nitride belongs to hexagonal symmetry. C-plane of gallium nitride has perfect three-fold rotation symmetry. Since the symmetry groups are different for GaN and sapphire, any other planes than C-plane of sapphire cannot grow a GaN crystal. Thus, the GaInN-LEDs in use include sets of c-axis grown InGaN, InGaAlN or GaN thin films grown on C-planes of sapphire substrates.
All the GaN or GaInN thin films heteroepitaxially grown on the sapphire substrates are C-plane growing crystals. Sapphire substrates cannot make non-C-plane growing GaN crystals at all. Since sapphire has been a unique seed crystal for growing GaN until recently, it has been impossible to make a non-C-plane GaN crystal. At present, all the GaInN-LEDs and GaInN-LDs on the market consist of a pile of C-plane grown GaN, InGaN or AlInGaN thin films grown on C-plane sapphire substrates.
Large lattice misfit between sapphire and gallium nitride induces plenty of dislocations in a gallium nitride crystal grown on a sapphire substrate. Gallium nitride has rigidity as high as ceramics. The rigidity maintains the framework of crystals for a long time. Inherent dislocations in GaN crystals of LEDs do not enlarge by current injection unlike GaAs crystals. Since the dislocations do not increase, the GaN crystals on sapphire do not degrade. In spite of the great many dislocations, GaN-LEDs enjoy a long life time, high reliability and good reputation.
Sapphire substrates, however, have some drawbacks. Sapphire is a very rigid, hard crystal. Sapphire lacks cleavage. Sapphire is an insulator. Rigidity, non-cleavage and insulation are weak points of sapphire. When a plenty of device units have been fabricated upon a sapphire substrate wafer by wafer processes, the device-carrying sapphire wafer cannot be divided by natural cleavage like silicon wafers. The sapphire wafer should be mechanically cut and divided into individual device chips. The mechanical dicing step raises the cost.
The non-cleavage is not a serious obstacle for making LEDs (light emitting diodes) on sapphire substrates, since an LED has no resonator mirror. In the case of producing LDs (laser diodes) on sapphire substrates, the non-cleavage is a fatal drawback. A laser diode (LD) requires two mirrors at both ends of an active (stripe) layer as a resonator for amplifying light by injected current. It is convenient to form resonator mirrors by natural cleavage in a laser diode, because natural cleaved planes are endowed with flatness and smoothness. On-sapphire LDs prohibit natural cleavage from making resonator mirrors. Flat, smooth mirrors should be made on both ends of the laser chips by a vapor phase etching method, e.g., RIE (reactive ion etching), which is a difficult operation. Mirror-polishing should be done chip by chip after the wafer process has finished. Mirror-finishing of the resonators by the RIE is a main reason raising the cost of manufacturing the on-sapphire GaInN-LDs.
Another drawback results from the fact that sapphire is an insulator. Insulation prevents on-sapphire LEDs and LDs from having an n-electrode on the bottom. Sapphire forces LEDs and LDs to have extra n-type layers upon an insulating substrate but below an active layer. The n-electrode is formed by partially etching away a p-layer and the active layer, revealing the extra n-layer and depositing an n-electrode alloy on the n-layer. Both a p-electrode and the n-electrode are formed on the top surface of the LED or LD. Since electric current flows in the horizontal direction, the n-layer should have a sufficient thickness. It takes much time to eliminate a part of the p-layer and form an ohmic n-electrode on the revealed n-layer. An increase of the steps and time enhances the cost of the on-sapphire LEDs. Both the n-electrode and the p-electrode occupy a wide area on the top of the LED, which raises a necessary area of the LED. On-sapphire GaInN-LEDs which are prevailing cannot conquer the above drawbacks yet.
A gallium nitride (GaN) single crystal substrate would be an ideal substrate which has a probability of solving the drawbacks. Since thin films of GaN or GaInN are epitaxially deposited upon a substrate for producing blue light LEDs and LDs, a GaN bulk single crystal would eliminate the problem of lattice misfitting between the deposited films and the substrate. If an n-type bulk single crystal GaN substrate can be produced, an n-electrode can be formed on the bottom of the n-type GaN substrate. An allocation of a p-electrode at the top and an n-electrode at the bottom facilitates to produce LEDs, to bond the LEDs on packages, and to wirebond the LEDs to wiring patterns on the packages. The bottom n-electrode enables an LED to reduce the chip size.
Another advantage results from cleavability of a bulk GaN single crystal. A device-produced GaN wafer can be divided into stripe arrays of individual device (LED or LD) chips by natural cleavage. However, cleavage planes (1-100), (01-10) and (−1010) are parallel to three sides of an equilateral triangle defined upon a C-plane (0001) of GaN. The GaN crystal has not a square set of cleavage planes but a triangle set of cleavage planes. Square device (LED or LD) chips are produced by cutting a device-carrying GaN wafer partially by natural cleavage and partially by mechanical dicing.
Furthermore, an LD (laser diode) chip can produce resonator mirrors by natural cleavage. Replacement of the RIE by the natural cleavage reduces the cost of making GaInN-type blue light laser diodes (LDs).
However, there is no mineral containing gallium nitride single crystals. No attempt of making a wide, bulk GaN single crystal substrate artificially has succeeded until recently. Since a GaN single crystal substrate was inaccessible, it was not possible to make GaInN type LEDs or LDs on a single crystal GaN substrate until recently.
Recently, vapor epitaxial methods which can grow a GaN single crystal on a foreign material substrate have been proposed and improved. The methods are described as follows.
(1) Metallorganic Chemical Vapor Deposition Method (MOCVD)
The most prevailing method for making GaN crystals is a Metallorganic Chemical Vapor Deposition Method (MOCVD). The MOCVD produces a GaN crystal by placing a sapphire substrate in a cold-wall furnace, heating the sapphire substrate, supplying a material gas including TMG (trimethylgallium) and ammonia (NH3) to the sapphire substrate, and synthesizing gallium nitride (GaN) from the material gas on the substrate. Although a great amount of the material gas is inhaled into the furnace, only a part of the material gas reacts with each other for making gallium nitride molecules. Other part of the material gas is dissipated in vain. The MOCVD is suffering from low yield and low growing speed. The MOCVD is favorable for making thin GaN films but is unsuitable for producing a thick GaN crystal layer due to the material dissipation. Another drawback is possibility of carbon contamination due to carbon included in the metallorganic gases.
(2) Metallorganic Chloride Method (MOC)
An MOC method produces a GaN crystal by placing a sapphire substrate or GaAs substrate in a hot-wall furnace, supplying TMG (trimethylgallium) and HCl (hydrochloric acid) into the furnace, synthesizing GaCl (gallium chloride) above the substrate, supplying ammonia (NH3) to the heated substrate, inducing a reaction between NH3 and GaCl on the substrate, making gallium nitride molecules on the substrate and depositing the gallium nitride molecules on the substrate. Since the MOC method makes once an intermediate compound GaCl, carbon contamination is alleviated in comparison with the MOCVD. However, the MOC is not fully immune from possibility of carbon contamination, since the MOC employs trimethylgallium gas.
(3) Hydride Vapor Phase Epitaxy Method (HVPE)
Unlike the MOCVD or the MOC, an HVPE employs metal Ga monoelement as a gallium source. FIG. 1 shows a HVPE apparatus having a hot-wall furnace 1. A heater 2 is upholstered around the furnace 1. Gas inlets 3 and 4 are provided at the top of the furnace 1 for introducing two kinds of material gases. The furnace 1 sustains a Ga-boat 5 at an upper space. A Ga-melt 6 is prepared by putting metal Ga into the Ga-boat 5 and heating the Ga-boat 5 by the heater 2. One gas inlet 3 has an open end facing to the Ga-boat 5 for supplying H2+HCl gas to the Ga-boat 5. The other gas inlet 4 has an open end at a middle height of the furnace for introducing H2+NH3 gas.
A susceptor 7 is supported by a rotation shaft 8 in a lower half of the furnace 1. The rotation shaft 8 can rotate, lift up or down the susceptor 7. A GaAs substrate or a GaN substrate 9 is laid upon the susceptor 7 as a seed. A GaN seed crystal can be prepared by making a GaN crystal on a GaAs substrate, eliminating the GaAs substrate and slicing the grown GaN crystal into wafers. The heater 2 heats the susceptor 7 and the substrate 9. An intermediate compound gallium chloride GaCl gas is synthesized by blowing the HCl+H2 gas to the Ga-melt 6 in the boat 5. GaCl falls in the furnace near the substrate 9, reacts with ammonia and synthesizes gallium nitride (GaN) on the substrate 9. The HVPE uses a non-carbon material (Ga monoelement). The HVPE is free from possibility of carbon contamination which degrades electric properties of object crystals.
(4) Sublimation Method
Heating alone cannot convert solid GaN into a melt of Ga. High pressure is required for melting solid GaN besides heating. Difficulty of making a GaN melt prohibits a Czochralski method or a Bridgman method from growing a GaN solid from a GaN melt. Without high pressure, solid GaN sublimes into vapor GaN by heating. A sublimation method makes a GaN single crystal on a substrate by inserting a substrate and a GaN polycrystal source into a reaction tube, heating the GaN polycrystal source for subliming at a higher temperature, heating the substrate at a lower temperature, transporting GaN vapor from the GaN source to the colder substrate and depositing GaN molecules on the substrate.
Another improvement (Lateral Overgrowth Method) has been proposed for making a low-dislocation density GaN film grown on a sapphire substrate for making on-sapphire GaInN-LEDs.
[Epitaxial Lateral Overgrowth Method (ELO)]
{circumflex over (1)} Akira Usui, “Thick Layer Growth of GaN by Hydride Vapor Phase Epitaxy”, Electronic Information and Communication Society, C-II, vol. J81-C-II, No. 1, pp 58-64, (January, 1998), proposed a growth of GaN by a lateral overgrowth method. The lateral overgrowth method produces a low dislocation density GaN crystal by covering a sapphire substrate with a mask having dotted or striped windows lying at corner points of periodically allocated equilateral triangles, supplying material gas via the mask windows to the sapphire substrate, depositing GaN molecules on the sapphire substrate within the windows, growing further GaN films from the windows over on the mask, joining neighboring GaN films in horizontal directions along boundaries between the windows on the mask, turning the growing direction from the horizontal directions to the vertical direction and maintaining the vertical GaN growth on the mask. Dislocations have a tendency of extending along the growing direction. Many dislocations accompany GaN growth in any cases. The ELO method forces the dislocations to bend at the meeting boundaries above the mask from horizontal extensions to an upward extension. The change of extension reduces the dislocations in the GaN crystal. Thus, the ELO is effective to reduce dislocation density of a GaN thin film grown on a sapphire substrate.
Inventors of the present invention chose the HVPE method as a very promising candidate among the mentioned vapor phase growth methods for growing a thick GaN crystal for a freestanding GaN wafer. Almost all of the preceding trials for growing GaN films have started from sapphire single crystals as a substrate. Sapphire has, however, some drawbacks of non-existence of cleavage and impossibility of removal. The inventors abandoned sapphire as a substrate for making a freestanding GaN single crystal.
Instead of sapphire, the inventors of the present invention chose GaAs (gallium arsenide) as a substrate for growing a thick GaN crystal for making a freestanding GaN single crystal wafer. Although GaAs has a cubic symmetry which is different from the hexagonal symmetry of GaN and the trigonal symmetry of sapphire, a (111) GaAs plane has three-fold rotation symmetry akin to the hexagonal symmetry. The inventors of the present invention succeeded in growing a GaN crystal on a (111) GaAs substrate in the c-direction from materials of metal gallium, hydrogen-diluted hydrochloric acid (HCl) gas and hydrogen-diluted ammonia gas. Fortunately, the GaAs substrate can be eliminated from the grown GaN crystal by etching or polishing. Possibility of removal is an advantage of GaAs as a substrate for making a freestanding GaN crystal.
{circumflex over (2)} Japanese Patent Application No. 10-183446(183446/'98) was filed by the same inventor as the present invention. {circumflex over (2)} produced a GaN single crystal by preparing a GaAs (111) substrate, covering the GaAs substrate with a mask having windows, growing a thick GaN layer on the masked GaAs substrate by the HVPE and the ELO method, and eliminating the GaAs substrate by aqua regia. {circumflex over (2)} obtained a freestanding GaN bulk single crystal wafer having a 20 mm diameter and a 0.07 mm thickness. The GaN crystal was a C-(0001) plane crystal.
{circumflex over (3)} Japanese Patent Application No. 10-171276(171276/'98) was filed by the same inventor as the present invention. {circumflex over (3)} also proposed a freestanding GaN bulk single crystal wafer of a C-plane produced by depositing a thick GaN crystal upon a (111) GaAs substrate. Distortion of the GaN wafer was a problem. Distortion is induced on the freestanding GaN wafer by differences of thermal expansion between GaAs and GaN. How to reduce the distortion was another problem for {circumflex over (3)}. Conduction type of the GaN crystal was left untouched.
{circumflex over (4)} Kensaku Motoki, Takuji Okahisa, Naoki Matsumoto, Masato Matsushima, Hiroya Kimura, Hitoshi Kasai, Kikurou Takemoto, Koji Uematsu, Tetsuya Hirano, Masahiro Nakayama, Seiji Nakahata, Masaki Ueno, Daijirou Hara, Yoshinao Kumagai, Akinori Koukitu and Hisashi Seki, “Preparation of Large Freestanding GaN Substrates by Hydride Vapor Phase Epitaxy Using GaAs as a Starting Substrate”, Jpn. J. Appl. Phys. Vol. 40 (2001) pp. L140-143, reported a freestanding GaN single crystal produced by a lateral overgrowth method upon a GaAs (111) substrate. Grown GaN was a (0001) C-plane crystal having a 500 μm thickness and a 2 inch diameter. The GaN crystal showed n-type conduction. Dislocation density was 2×105 cm−2. Carrier density was n=5×1018 cm−3. Mobility was 170 cm2/Vs. Resistivity was 8.5×10−3 Ωcm. {circumflex over (4)} said nothing about n-dopants.
{circumflex over (5)} Japanese Patent Application No. 11-144151 was filed by the same inventor as the present invention. {circumflex over (5)} proposed a freestanding n-type GaN single crystal containing oxygen as an n-dopant having nearly 100% of activation rate in GaN. This was the first document which asserted that oxygen was a good n-dopant in GaN with nearly a 100% activation rate. Silicon (Si) has been prevailing as an n-dopant which has been exclusively doped into GaN thin films grown on sapphire substrates in a form of silane gas (SiH4). But silane gas (SiH4) is a dangerous gas. Oxygen can be supplied in a safe form of water or water vapor to material gases. {circumflex over (5)} rejected silicon but admitted oxygen as an n-dopant in GaN. {circumflex over (5)} insisted on replacement of silicon by oxygen. {circumflex over (5)} alleged that carbon (C) which is an n-impurity and disturbs the action of oxygen should be excluded from the material gases. {circumflex over (5)} denied the MOCVD (metallorganic chemical vapor deposition) method which uses metallorganic gases including plenty of carbon atoms but recommended the HVPE (hydride vapor phase epitaxy) method.
GaN is a hexagonal symmetry crystal with three-fold rotational symmetry. Crystallographical representation of GaN is different from GaAs(zinc blende type) which belongs to the cubic symmetry group. Crystallographic representation of the hexagonal symmetry group is now described. There are two representations for hexagonal symmetry. One method uses three parameters. The other method uses four parameters. Here, four parameter representation which requires four axes is described. Three axes are denoted by a-axis, b-axis and d-axis which lie on the same horizontal plane and meet at an origin with each other at 120 degrees. Unit lengths a, b and d of the three axes are equal, that is, a=b=d.
An extra axis meets with other three axes at 90 degrees. The extra axis is named c-axis. A set of a-, b-, d-, and c-axes defines planes and directions in a hexagonal symmetry crystal. The three a-, b-, d-axes are equivalent. But the c-axis is a unique axis. A set of plenty of parallel equivalent planes is imagined. When a first plane crosses a-axis at a point of a/h, b-axis at a point of b/k, d-axis at a point of d/m and c-axis at a point of c/n, the plane is represented by (hkmn). When the first plane cannot cross a positive part of the axes, the axes should be extended in a negative direction for crossing with the first plane. Crystal has a periodic character. Thus, h, k, m and n are positive or negative integers including zero (0). Number “h” means the number of the object planes existing in a unit length “a” of a-axis. Number “k” means the number of the object planes existing in a unit length “b” of b-axis. The object plane is represented a round bracketed indices (hkmn).
Three equivalent indices h, k and m for a-, b-, d-axes always satisfy a zero-sum rule of h+k+m=0. The other index n for c-axis is a free parameter. Crystal indices h, k, m and n are substituted in a bracket without comma “,”. A negative index should be discriminated from a positive one by upperlining by the crystallograpy. Since an upperline is forbidden, a negative index is shown by adding a minus sign before the integer. There are two index representations. One is an individual representation. The other is a collective representation. Objects of the index representation are planes and directions. A direction and a plane take the same set hkmn of indices, when the direction is a normal (meeting at 90 degrees) to the plane. But the kinds of brackets are different.
Round bracketed (hkmn) means an individual representation of a plane. Wavy bracketed {hkmn} means a collective representation of family planes. Family planes are defined as a set of planes all of which can be converted into other member planes by the symmetry operation included in the crystal symmetry.
Besides the plane representation, linear directions should be denoted by a similar manner. Square bracketed [hkmn] means an individual representation of a direction which is vertical to an individual plane (hkmn). Edged bracketed <hkmn> means a collective representation of family directions. Family directions are defined as a set of directions all of which can be converted into other member directions by the symmetry operation of the crystal symmetry. The definitions are shown as follows.
(hkmn)individual, plane.{hkmn}collective, plane.[hkmn]individual, direction.<hkmn>collective, direction.
In the hexagonal symmetry, C-plane is the most important plane represented as (0001) which is normal to the horizontal plane including a-, b- and c-axes. C-plane has three-fold rotational symmetry. All of the artificially made GaN crystals have been produced by C-plane growth which grows a crystal by maintaining C-plane as a surface. When GaN is heteroepitaxially grown on a foreign material, for example, sapphire (Al2O3) or gallium arsenide (GaAs), the seed surface should have three-fold rotational symmetry. Thus, GaN grows on the foreign substrate with the three-fold rotation symmetry, maintaining C-plane which has also the same symmetry. Thus, heteroepitaxy on a foreign substrate is restricted to C-plane growth. There are two secondary important planes next to C-plane.
One important plane is {1-100} planes which are vertical to C-plane. This is a cleavage plane. The {1-100} planes mean a set of six individual planes (1-100), (10-10), (01-10), (−1100), (−1010) and (0-110) which are all cleavage planes. The (1-100), (10-10), (01-10), (−1100), (−1010) and (0-110) planes are called “M-plane” for short. The cleavage planes meet with each other at 60 degrees. Any two cleavage planes are not vertical.
The other important plane is {11-20} planes which are vertical to C-plane. The {11-20} planes mean a set of six individual planes (11-20), (1-210), (−2110), (2-1-10), (−12-10) and (−1-120). The (11-20), (1-210), (−2110), (2-1-10), (−12-10) and (−1-120) planes are called “A-plane” for short. A-planes are not cleavage planes. A-planes meet with each other at 60 degrees.
C-plane {0001} is uniquely determined. But A-planes and M-planes are not uniquely determined, since A-planes and M-planes include three different planes. Some of A-planes are vertical to some of M-planes.
All A-planes are vertical to C-plane. All M-planes are vertical to C-plane. Some of A-planes, some of M-planes and C-plane can build a set of orthogonal planes.
{circumflex over (6)} Japanese Patent Application No. 10-147049(147049/'98) was filed by the same inventor as the present invention. {circumflex over (6)} proposed non-rectangular GaN devices which have sides of cleavage planes (M-planes). The GaN crystal has C-plane as a surface.
{circumflex over (7)} Japanese Patent Application No. 11-273882(273882/'99) was filed by the same inventor as the present invention. Conventional growth means a growth by maintaining a mirror-flat, even C-plane surface. {circumflex over (7)} proposed facet-growth of GaN along c-axis which keeps various facets on C-plane. The facets on C-plane signify small other planes than C-plane. Facets form hexagonal pits or hillocks and dodecagonal pits or hillocks on C-plane. Although GaN grows on an average along the c-axis, various facets cover a surface of growing GaN. The facets sweep dislocations down to the bottoms of the facet pits. Dislocations are effectively reduced by the facets.
{circumflex over (8)} Japanese Patent Application No. 2000-207783(207783/'00) was filed by the same inventor as the present invention. {circumflex over (8)} discovered a fact that dislocations extend in parallel with the growing direction in a GaN crystal. C-plane growth prolongs dislocations in parallel with the c-axis. {circumflex over (8)} proposed a sophisticated method of growing a tall GaN crystal in the c-direction on a C-plane of a GaN seed, cutting the GaN crystal in A planes, obtaining an A-plane GaN seed crystal, growing the GaN crystal in an A-direction on the A-plane seed, cutting the A-grown GaN in M planes and obtaining M-plane GaN seeds with low dislocation density. Prior art of {circumflex over (1)} to {circumflex over (7)} grow C-plane GaN crystals in the c-direction on a foreign material or a C-plane GaN seed. Only {circumflex over (8)} proposed non-C-plane growth of GaN on a non-C-plane GaN substrate.
All of the known attempts of on-sapphire GaN growth grow C-surface GaN crystals having a C-surface as a top without exception. There is a reason of the absolute prevalence of C-plane GaN crystals. When a sapphire (α-Al2O3) single crystal substrate is used as a substrate, a C(0001) surface sapphire is used to be chosen. Sapphire belongs to a trigonal symmetry group which requires four indices for representing orientations of planes and directions. GaN has hexagonal symmetry. Lengths of c-axes are different between sapphire and gallium nitride. GaN has three typical, low index planes of A-plane, M-plane and C-plane, as explained before. A-plane or M-plane GaN cannot grow on A-plane or M-plane sapphire, because A-plane or M-plane are too complex to coincide with a similar plane of sapphire without misfit. Only C-plane of GaN can join on C-plane of sapphire. Thus, a smooth, flat C-plane GaN crystal can be easily grown on a C-plane sapphire substrate. All of the known InGaN-LEDs have a pile of C-plane GaN or InGaN layers on a C-plane sapphire substrate.
Similarly, in the case of GaAs substrates, a three-fold rotationally symmetric (111) GaAs substrate is selected as a substrate. GaAs belongs to a cubic symmetry group. But a (111) plane of GaAs has three-fold rotational symmetry. The (111) GaAs substrate allows only C-plane GaN having the corresponding rotation symmetry to grow on. Foreign materials as a substrate cannot grow non-C-plane GaN at all.
A thick bulk crystal requires far more amount of dopants than a thin film crystal. The amount of dopants is in proportion to a thickness or volume of grown crystals. Silane gas (SiH4) is a dangerous gas which sometimes induces a burst. An n-type GaN bulk crystal would require a larger amount of an n-dopant than an n-type thin film. The inventors prefer oxygen to silicon as the n-dopant, because water (H2O) as an oxygen compound is far safer than silane gas (SiH4). The inventors tried to dope GaN bulk crystals with oxygen as an n-dopant. However, mirror flat, C-plane GaN crystals cannot be doped enough with oxygen. The inventors discovered a fact that oxygen does not go into GaN easily via C-plane and C-plane repulses oxygen. The inventors found orientation dependence of oxygen doping for the first time. The inventors were aware of the fact that oxygen doping has orientation dependence. The inventors noticed that C-plane is the worst plane for oxygen doping.
The orientation dependence of oxygen doping is a novel phenomenon. It is not easy to understand the orientation dependence of oxygen doping. Nobody found the phenomenon before the inventors of the present invention. The inventors analysed atomic components on a surface of C-plane grown GaN crystals by SIMS (Secondary Ion Mass Spectroscopy). The SIMS determines atomic ratios on a surface of an object sample by emitting first ions, accelerating the first ions, shooting the first ions at the sample for inducing secondary ions emitted out of the sample, analysing the mass of the secondary ions and counting the numbers of the emitted secondary ions. The numbers of the emitted secondary ions are proportional to products of emission efficiencies of secondary ions and atomic ratios on the object surface. Since the emission efficiencies are known parameters, the atomic ratios are determined.
At an early stage of the SIMS analysis, insufficient resolution of the secondary ions and broadness of the first ion beam allowed a wide secondary beam to emanate from a wide area of the object. The broad secondary ion beams obscured abnormality of oxygen doping. Then, C-planes of GaN seemed to emit oxygen secondary ions.
At a later stage, the inventors succeeded in converging the first ion beam and enhancing resolution of the SIMS. Narrow converged first ions and enhanced resolution revealed a surprising fact.
A C-plane surface of a C-grown GaN includes microscopic pits or hillocks. The microscopic hillocks and pits have many non-C-planes which are called facets. A rough C-plane is an assembly of micro C-planes and micro facets. Oxygen secondary ions were measured by discriminating the micro C-planes and micro non-C-plane facets. The inventors found that the oxygen secondary ions were emitted from the micro non-C-plane facets and the micro C-planes do not emit the oxygen secondary ions. Namely, the micro C-planes included far smaller rate of oxygen than an average rate. When oxygen concentration was 5×1018 cm−3 at non-C-plane facets, oxygen concentration was less than 1×1017 cm−3 at C-planes. The facets have oxygen acceptance function which is more than 50 times as large as that of C-planes. C-planes are the poorest in the function of accepting oxygen. In the SIMS experiments, the secondary oxygen ions emanated not from C-planes but from the facets.
The inventors made a mirror-flat C-plane GaN crystal. The oxygen concentration was less than 1×1017 cm−3 everywhere on a sample.
The SIMS analysis taught us that oxygen is hardly doped into C-plane of GaN. The oxygen doping to a C-plane grown GaN crystal is caused by the facets which have high acceptance function of oxygen.
When a GaN crystal grows in an average in the c-direction, oxygen can be doped into GaN via the facets. Facets enable C-plane GaN to accept oxygen as an n-dopant. The oxygen accepting power is in proportion to the area of the facets. The wider the facets are, the stronger the oxygen acceptance power is. When C-plane is covered overall with facets, the oxygen acceptance power attains to the maximum.
Otherwise, an A-plane GaN seed and an M-plane GaN seed are best substrates for growing a non-C-plane GaN crystal and doping the non-C-plane GaN crystal with oxygen.
In conclusion, Si which is a prevalent n-dopant for thin films of GaN is not suitable for doping large GaN bulk crystals. Oxygen is a safer n-dopant. But conventional C-plane growth repulses oxygen. Oxygen doping is insufficient for the C-plane growth. An A-plane GaN seed, an M-plane seed and facet c-axis growth are effective to oxygen doping into GaN crystals.